Exploiting Underlying Structure for Detailed Reconstruction of an Internet-scale Event

Transcription

1 Exploiting Underlying Structure for Detailed Reconstruction of an Internet-scale Event Abhishek Kumar (Georgia Tech / Google) Vern Paxson (ICSI) Nicholas Weaver (ICSI) Proc. ACM Internet Measurement Conference

2 Enhancing Telescope Imagery NGC6543: Chandra X-ray Observatory Center ( 2

3 Enhancing Telescope Imagery NGC6543: Chandra X-ray Observatory Center ( 3

4 The Witty Worm Released March 19, Exploited flaw in the passive analysis of Internet Security Systems products Worm fit in a single Internet packet Stateless: When scanning, worm could fire and forget Vulnerable pop. (12K) attained in 75 minutes. Payload: slowly corrupt random disk blocks. Flaw had been announced the previous day. Written by a Pro. 4

5 What Exactly Does Witty Do? 1. Seed the PRNG using system uptime. 2. Send 20,000 copies of self to randomly selected destinations. 3. Open physical disk chosen randomly between If success: 5. Overwrite a randomly chosen block on this disk. 6. Goto line Else: 8. Goto line 2. 5

6 Witty Telescope Data UCSD telescope recorded every Witty packet seen on /8 (2 24 addresses). But with unknown losses In the best case, we see 4 of every 1,000 packets sent by each Witty infectee.? What can we figure out about the worm? 6

7 Generating (Pseudo-)Random Numbers Linear Congruential Generator (LCG) proposed by Lehmer, 1948: X i+1 = X i *A + B mod M Picking A, B takes care, e.g.: A = 214,013 B = 2,531,011 M = 2 32 Theorem: the orbit generated by these is a complete permutation of Another theorem: we can invert this generator 7

8 srand(seed) { X seed } rand() { X X* ; return X } main() 1. srand(get_tick_count()); 2. for(i=0;i<20,000;i++) 3. dest_ip rand() [0..15] rand() [0..15] 4. dest_port rand() [0..15] 5. packetsize rand() [0..8] 6. packetcontents top-of-stack 7. sendto() 8. if(open_physical_disk(rand() [13..15] )) 9. write(rand() [0..14] 0x4e20) 10. goto else goto 2 8

9 What Can We Do Seeing Just 4 Packets Per Thousand? Each packet contains bits from 4 consecutive PRNGs: 3. dest_ip rand() [0..15] rand() [0..15] 4. dest_port rand() [0..15] 5. packetsize rand() [0..8] If first call to rand() returns X i : 3. dest_ip (X i ) [0..15] (X I+1 ) [0..15] 4. dest_port (X I+2 ) [0..15] Given top 16 bits of X i, now brute force all possible lower 16 bits to find which yield consistent top 16 bits for X I+1 & X I+2 Single Witty packet suffices to extract infectee s complete PRNG state! Think of this as a sequence number. 9

10 Cool, But So What? E.g., Individual Access Bandwidth Estimation Suppose two consecutively-observed packets from source S arrive with states X i and X j Compute j-i by counting # of cranks forward from X i to reach X j # packets sent between the two observed = (j-i)/4 sendto call in Windows is blocking Ergo, access bandwidth of that infectee should be (j-i)/4 * size-of-those-packets / ΔT Note: works even in the presence of very heavy packet loss 10

11 Inferred Access Bandwidth of Individual Witty Infectees 11

12 Precise Bandwidth Estimation vs. Rates Measured by Telescope 12

13 Systematic Telescope Loss 13

14 Telescope Comparison 14

15 Telescope Bias 15

16 srand(seed) { X seed } rand() { X X* ; return X } main() 1. srand(get_tick_count()); 2. for(i=0;i<20,000;i++) 3. dest_ip rand() [0..15] rand() [0..15] 4. dest_port rand() [0..15] 5. packetsize rand() [0..8] 6. packetcontents top-of-stack 7. sendto() 8. if(open_physical_disk(rand() [13..15] )) 9. write(rand() [0..14] 0x4e20) 10. goto else goto 2 } } Or complete reseeding if not }4 calls to rand() per loop Plus one more every 20,000 packets, if disk open fails 16

17 Witty Infectee Reseeding Events For packets with state X i and X j : If from the same batch of 20,000 then j - i = 0 mod 4 If from separate but adjacent batches, for which Witty did not reseed, then j - i = 1 mod 4 (but which of the 100s/1000s of intervening packets marked the phase shift?) If from batches across which Witty reseeded, then no apparent relationship. 17

18 18

19 19

20 20

21 21

22 22

23 23

24 24

25 25

26 26

27 27

28 28

29 29

30 30

31 We Know Intervals in Which Each First-Seed Packet Occurs. but which among the 1,000s of candidates are the actual seeds? Entropy isn t all that easy to come by Consider srand(get_tick_count()) i.e., uptime in msec The values used in repeated calls increase linearly with time 31

32 32

33 33

34 34

35 35

36 36

37 37

38 Slope = 1000/sec Time back to X-intercept = uptime 38

39 Uptime of 750 Witty Infectees? 39

40 Uptime of 750 Witty Infectees 40

41 Given Exact Values of Seeds Used for Reseeding we know exact random # used at each subsequent disk-wipe test: if(open_physical_disk(rand() [13..15] ) and its success, or failure, i.e., number of drives attached to each infectee and, more, generally, every packet each infectee sent Can compare this to when new infectees show up i.e. Who-Infected-Whom 41

42 Disk Drives Per Witty Infectee ? % Infectees w/ # Drives 42

43 Disk Drives Per Witty Infectee % Infectees w/ # Drives

44 Given Exact Values of Seeds Used for Reseeding we know exact random # used at each subsequent disk-wipe test: if(open_physical_disk(rand() [13..15] ) and its success, or failure, i.e., number of drives attached to each infectee and, more, generally, every packet each infectee sent Can compare this to when new infectees show up i.e. Who-Infected-Whom 44

45 Time Between Scan by Known Infectee and New Source Arrival At Telescope Too Early Too Late Right on Time 45

46 Infection Attempts That Were Too Early, Too Late, or Just Right Infector/Infectee Signature 46

47 Witty is Incomplete Recall that LCD PRNG generates a complete orbit over a permutation of But: Witty author didn t use all 32 bits of single PRNG value dest_ip (X i ) [0..15] (X I+1 ) [0..15] Knuth recommends top bits as having better pseudo-random properties But 2 : This does not generate a complete orbit! Misses 10% of the address space Visits 10% of the addresses (exactly) twice So, were 10% of the potential infectees protected? 47

48 Time When Infectees Seen At Telescope Doubly-scanned infectees infected faster Unscanned infectees still get infected! In fact, some are infected Extremely Quickly! 48

49 How Can an Unscanned Infectee Become Infected? Multihomed host infected via another address Might show up with normal speed, but not early DHCP or NAT aliasing Would show up late, certainly not early Could they have been passively infected extra quickly because they had large crosssections? Just what are those hosts, anyway? 49

50 Uptime of 750 Witty Infectees Part of a group of 135 infectees from same /16 50

51 Time When Infectees Seen At Telescope Most also belong to that /16 51

52 Analysis of the Extra-Quick Hosts Initial infectees exhibit super-exponential growth they weren t found by random scanning Hosts in prevalent /16 numbered x.y.z.4 in consecutive /24 subnets Lineage analysis reveals that these subnets not sufficiently visited at onset to account for infection One possibility: they monitored networks separate from their own subnet But: if so, strange to number each.4 in adjacent subnets Unlikely infection was due to passive monitoring 52

53 Alternative: Witty Started With A Hit List Unlikely infection was due to passive monitoring Prevalent /16 = U.S. military base Attacker knew of ISS security software installation at military site ISS insider (or ex-insider) Fits with very rapid development of worm after public vulnerability disclosure 53

54 Are All The Worms In Fact Executing Witty? Answer: No. There is one infectee that probes addresses not on the orbit. Each probe contains Witty contagion, but lacks randomized payload size. Shows up very near beginning of trace. Patient Zero - machine attacker used to launch Witty. (Really, Patient Negative One.) European retail ISP. Information passed along to Law Enforcement. 54

55 Summary of Witty Telescope Forensics Understanding a measurement s underlying structure adds enormous analytic power Cuts both ways: makes anonymization much harder than one would think With enough effort, worm attribution can be possible But a lot of work And no guarantee of success 55